You Can’t Take a Random Sample from the Future by Dan Strongin

How some­thing that knows it is be­ing mea­sured will dis­tort the pro­cess or the data when given an un­re­al­is­tic tar­get, and how dis­tor­tion re­lates to what hap­pened with Ex­ploitGym.Post one in the se­ries ˜When Mea­sur­ing Makes Things Worse˜. It looked at dis­tor­tion. Then it cov­ered tam­per­ing: re­act­ing to or­di­nary vari­a­tion as if it means some­thing. Now I go into a rule to sep­a­rate or­di­nary from a sig­nal and what fol­lows from that.Note: Th­ese words are mine. I wrot...